Update. cvs/libc-ud-970813

1997-08-14 03:14  Ulrich Drepper  <drepper@cygnus.com>

	* elf/dl-minimal.c: Don't use relative #include paths.
	(malloc): Don't try to initialize _dl_pagesize, it already is.
	(__sigjmp_save, longjmp): Reformat.

	* sysdeps/generic/dl-sysdep.c: Implement _dl_show_auxv function to
	show content of auxiliary array.
	(_dl_sysdep_start): Remember start of auxiliary vector.
	* sysdeps/mach/hurd/dl-sysdep.c: (_dl_show_auxv): New dummy function.
	* elf/link.h: Add prototype for _dl_show_auxv.
	* elf/rtld.c (dl_main): Call _dl_show_auxv if LD_SHOW_AUXV envvar
	is present.

	* libio/iofdopen.c: Better prepare for use in libstdc++.
	* libio/iofflush.c: Likewise.
	* libio/iofgetpos.c: Likewise.
	* libio/iofgets.c: Likewise.
	* libio/iofopen.c: Likewise.
	* libio/iofputs.c: Likewise.
	* libio/iofread.c: Likewise.
	* libio/iofsetpos.c: Likewise.
	* libio/ioftell.c: Likewise.
	* libio/iofwrite.c: Likewise.
	* libio/iogetdelim.c: Likewise.
	* libio/iogets.c: Likewise.
	* libio/iopopen.c: Likewise.
	* libio/ioputs.c: Likewise.
	* libio/iosetbuffer.c: Likewise.
	* libio/iosetvbuf.c: Likewise.
	* libio/ioungetc.c: Likewise.
	* libio/iovsprintf.c: Likewise.
	* libio/iovsscanf.c: Likewise.

	* manual/creature.texi: Mention _XOPEN_SOURCE_EXTENDED in correct
	place.

	* manual/math.texi: Add comment to sincos decsription to say it's a
	GNU extension.
	Fix description of random function.

	* posix/fnmatch.h: Recognize Windog by _WINDOWS32 not WIN32.
	* posix/glob.c: Likewise.  Handle stupid system headers on SunOS.
	Add casts for all __alloca calls.
	* posix/glob.h: Recognize Windog by _WINDOWS32 not WIN32.
	Declare additional GLOB_* constants also if _GNU_SOURCE is defined.
	Patches by Paul D. Smith <psmith@BayNetworks.COM>.

	* stdio-common/_itoa.c: Don't use relateive #include paths.

	* sysdeps/i386/fpu/bits/mathinline.h: Enable optimizations even for
	gcc 2.7.
	(atan): Loading 1.0 from memory does not pay off.
	(asinh): Fix typo.

1997-08-13 09:44  Philip Blundell  <Philip.Blundell@pobox.com>

	* sysdeps/stub/accept.c: Use socklen_t where necessary.
	* sysdeps/stub/bind.c: Likewise.
	* sysdeps/stub/connect.c: Likewise.
	* sysdeps/stub/getsockname.c: Likewise.
	* sysdeps/stub/getsockopt.c: Likewise.
	* sysdeps/stub/recvfrom.c: Likewise.
	* sysdeps/stub/sendto.c: Likewise.
	* sysdeps/stub/setsockopt.c: Likewise.

1997-08-11  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* io/ftw.c (object_compare): Avoid memcmp when comparing scalar
	types.

1997-08-12  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* math/libm-test.c (cbrt_test): Add epsilon for long double to
	`cbrt (-0.001)'.

1997-08-11  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/stub/bits/string.h: Fix comment.

1997-08-01  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* elf/Makefile (routines): Remove dl-version.
	(dl-routines): Add it here instead.  This removes dead code from
	libc.so.

1997-08-12 17:17  Wolfram Gloger  <wmglo@dent.med.uni-muenchen.de>

	* malloc/thread-m.h (thread_atfork, thread_atfork_static): New
	macros to encapsulate pthread_atfork's functionality.
	* malloc/malloc.c (ptmalloc_lock_all, ptmalloc_unlock_all): New
	functions, registered via thread_atfork in the initialization
	routine or via thread_atfork_static in global scope.  This
	ensures a consistent state of all locks across fork().

1997-08-11 23:00:32 1997  Mark Kettenis  <kettenis@phys.uva.nl>

	* login/programs/xtmp.c (compare_entry):
	Partially undo patch made by Klaus Espenlaub.  There
	was no xtmp/utmp typo.  Added comment to cclarify this.
	* login/programs/xtmp.h: Likewise.

	* posix/getopt.c: Recognize Windog by _WINDOWS32 not WIN32.
	Patch by Paul D. Smith <psmith@BayNetworks.COM>.

1997-08-12 13:51  Ulrich Drepper  <drepper@cygnus.com>

	* login/getutid.c: Rename to __getutid and make getutid and getutxid
	weak aliases.
	Patch by ir. Mark M._Kettenis <kettenis@phys.uva.nl>.

1997-08-11 23:55  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/unix/sysv/linux/Makefile [subdir=misc] (sysdep_headers):
	Add bits/mman.h.

1997-08-11 08:51:21 1997  H.J. Lu  <hjl@gnu.ai.mit.edu>

	* stdlib/fmtmsg.c (fmtmsg): Add static to
	__libc_once_define macro.

	* libc.map (__progname_full, __progname): Added.

1997-08-04 08:31  H.J. Lu  <hjl@gnu.ai.mit.edu>

	* sysdeps/unix/sysv/linux/if_index.c (if_nametoindex): Return 0
	if SIOGIFINDEX is not defined.
	(if_indextoname): Initialize "result" to NULL.
	(if_nametoindex, if_nameindex): Fix ENOSYS setting.

1997-08-11 11:58  Richard Henderson  <rth@cygnus.com>

	Sparc ISO C 9x exception handling:
	* sysdeps/sparc/fpu/bits/mathdef.h: New file.
	* sysdeps/sparc/fpu/fclrexcpt.c: New file.
	* sysdeps/sparc/fpu/fegetenv.c: New file.
	* sysdeps/sparc/fpu/fegetround.c: New file.
	* sysdeps/sparc/fpu/feholdexcpt.c: New file.
	* sysdeps/sparc/fpu/fesetenv.c: New file.
	* sysdeps/sparc/fpu/fesetround.c: New file.
	* sysdeps/sparc/fpu/feupdateenv.c: New file.
	* sysdeps/sparc/fpu/fgetexcptflg.c: New file.
	* sysdeps/sparc/fpu/fraiseexcpt.c: New file.
	* sysdeps/sparc/fpu/fsetexcptflg.c: New file.
	* sysdeps/sparc/fpu/ftestexcept.c: New file.
	* sysdeps/sparc/sparc32/fpu/bits/fenv.h: New file.
	* sysdeps/sparc/sparc64/fpu/bits/fenv.h: New file.

1997-08-11 11:58  Richard Henderson  <rth@cygnus.com>

	Sparc64 merge:
	* configure.in: Change up subdirectories, rename sparc8->sparcv8.

	* csu/Makefile: Nuke need-nopic-initfini bits.  The problem was
	that sparc64 ld did not recognize pic relocs at all.

	* elf/dl-deps.c: Include <string.h>.
	* elf/do-rel.h (elf_dynamic_do_rel): Take reloc address and size
	directly instead of the DT names to get them from.
	* elf/dynamic-link.h (_ELF_DYNAMIC_DO_RELOC): New macro distilled
	from ELF_DYNAMIC_DO_REL{,A}.  Define a second version that accounts
	for the possibility that DT_RELA and DT_JMPPLT overlap, as such is
	the case on the Sparc.

	* elf/elf.h: Stub out Elf64_Ver*, and make them the same as the
	Elf32_Ver* bits.  The linker does not even know to differentiate
	between the word sizes.  Someone should examine this and figure
	out what the proper sizes of things should be for 64-bit hosts
	before the sparc64 libraries are distributed.

	* shlib-versions: Rename sparc64 ld.so to not conflict with sparc32
	on the same machine.

	Move old sysdeps/sparc contents to sysdeps/sparc/sparc32,
	rename sparc8 subdir to sparcv8, move sysdeps/sparc64 content to
	sysdeps/sparc/sparc64.  Same for sysdeps/unix/sysv/linux/sparc.

	* sysdeps/sparc/Implies: Do ieee754 here.
	* sysdeps/sparc/configure.in: Move from sparc32/.
	* sysdeps/sparc/sparc32/Implies: Do wordsize here.
	* sysdeps/sparc/sparc64/Implies: Likewise.

	* sysdeps/sparc/sparc32/dl-machine.h (elf_machine_load_address):
	Use uninitialized .got to find load address instead of AT_BASE, as
	the later does not exist when running ld.so directly.

	* sysdeps/sparc/sparc32/fpu/fpu_control.h: Move from ../.
	* sysdeps/sparc/sparc64/fpu/fpu_control.h: Likewise.

	* sysdeps/sparc/sparc32/udiv_qrnnd.S: Clean up PIC support.

	* sysdeps/sparc/sparc64/Makefile (sysdep_routines): Add s_finitel
	and s_signbitl for stdio.
	* sysdeps/sparc/sparc64/configure.in: Remove, we no longer need
	nopic_initfini.

	* sysdeps/sparc/sparc64/bsd-_setjmp.S: Remove.
	* sysdeps/sparc/sparc64/bsd-setjmp.S: Remove.
	* sysdeps/unix/sysv/linux/sparc/sparc64/bsd-_setjmp.S: New empty file.
	* sysdeps/unix/sysv/linux/sparc/sparc64/bsd-setjmp.S: New empty file.
	* sysdeps/unix/sysv/linux/sparc/sparc64/longjmp.S: Add __longjmp alias.
	* sysdeps/unix/sysv/linux/sparc/sparc64/setjmp.S: Add weak _setjmp
	and setjmp entry points.

	* sysdeps/sparc/sparc64/dl-machine.h: Finish real implementation.

	* sysdeps/sparc/sparc64/elf/crtbegin.S: Add copyright, fix PIC.
	* sysdeps/sparc/sparc64/elf/crtend.S: Likewise.

	* sysdeps/unix/sysv/linux/sparc/bits/sigaction.h (struct sigaction):
	Fix sa_flags size for sparc64; add sa_restorer for kernel.

	* sysdeps/unix/sysv/linux/sparc/sparc64/brk.S: Fix PIC.
	* sysdeps/unix/sysv/linux/sparc/sparc64/clone.S: Likewise.
	* sysdeps/unix/sysv/linux/sparc/sparc64/init-first.h: Likewise.
	* sysdeps/unix/sysv/linux/sparc/sparc64/sysdep.h: Likewise.

	* sysdeps/unix/sysv/linux/sparc/sparc64/syscall.S: New file.

1997-08-10 18:29:08 1997  Mark Kettenis  <kettenis@phys.uva.nl>

	* login/login.c (login): Check for correct return value of
	utmpname.

1997-08-11 16:49  Ulrich Drepper  <drepper@cygnus.com>

	* time/tzset.c (__tzset_internal): Correctly handle TZ strings
	following incorrect old POSIX specs.
	Patch by Paul Eggert <eggert@twinsun.com>.

	* sysdeps/generic/bits/sigaction.h: Correct comment for SA_RESTART.
	* sysdeps/unix/bsd/osf/bits/sigaction.h: Likewise.
	* sysdeps/unix/sysv/linux/alpha/bits/sigaction.h: Likewise.
	* sysdeps/unix/sysv/linux/bits/sigaction.h: Likewise.
	* sysdeps/unix/sysv/linux/mips/bits/sigaction.h: Likewise.
	* sysdeps/unix/sysv/linux/sparc/bits/sigaction.h: Likewise.
	* sysdeps/unix/sysv/minix/bits/sigaction.h: Likewise.
	* sysdeps/unix/sysv/sysv4/bits/sigaction.h: Likewise.
	Reported by NIIBE Yutaka <gniibe@mri.co.jp>.

1997-08-09  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* manual/math.texi: Define the macros @mul and @infinity and use
	them to format the multiplication and infinity signs.
	(FP Comparison Functions): Use @code, not @math, for C code
	examples.

	* elf/dl-profile.c: Add descriptions and implement reading of

1 files changed, 153 insertions, 0 deletions

diff --git a/sysdeps/sparc/sparc32/umul.S b/sysdeps/sparc/sparc32/umul.S
new file mode 100644
index 0000000000..7a26c295cb
--- /dev/null
+++ b/sysdeps/sparc/sparc32/umul.S
@@ -0,0 +1,153 @@
+/*
+ * Unsigned multiply.  Returns %o0 * %o1 in %o1%o0 (i.e., %o1 holds the
+ * upper 32 bits of the 64-bit product).
+ *
+ * This code optimizes short (less than 13-bit) multiplies.  Short
+ * multiplies require 25 instruction cycles, and long ones require
+ * 45 instruction cycles.
+ *
+ * On return, overflow has occurred (%o1 is not zero) if and only if
+ * the Z condition code is clear, allowing, e.g., the following:
+ *
+ *	call	.umul
+ *	nop
+ *	bnz	overflow	(or tnz)
+ */
+
+#include "DEFS.h"
+FUNC(.umul)
+	or	%o0, %o1, %o4
+	mov	%o0, %y		! multiplier -> Y
+	andncc	%o4, 0xfff, %g0	! test bits 12..31 of *both* args
+	be	Lmul_shortway	! if zero, can do it the short way
+	andcc	%g0, %g0, %o4	! zero the partial product and clear N and V
+
+	/*
+	 * Long multiply.  32 steps, followed by a final shift step.
+	 */
+	mulscc	%o4, %o1, %o4	! 1
+	mulscc	%o4, %o1, %o4	! 2
+	mulscc	%o4, %o1, %o4	! 3
+	mulscc	%o4, %o1, %o4	! 4
+	mulscc	%o4, %o1, %o4	! 5
+	mulscc	%o4, %o1, %o4	! 6
+	mulscc	%o4, %o1, %o4	! 7
+	mulscc	%o4, %o1, %o4	! 8
+	mulscc	%o4, %o1, %o4	! 9
+	mulscc	%o4, %o1, %o4	! 10
+	mulscc	%o4, %o1, %o4	! 11
+	mulscc	%o4, %o1, %o4	! 12
+	mulscc	%o4, %o1, %o4	! 13
+	mulscc	%o4, %o1, %o4	! 14
+	mulscc	%o4, %o1, %o4	! 15
+	mulscc	%o4, %o1, %o4	! 16
+	mulscc	%o4, %o1, %o4	! 17
+	mulscc	%o4, %o1, %o4	! 18
+	mulscc	%o4, %o1, %o4	! 19
+	mulscc	%o4, %o1, %o4	! 20
+	mulscc	%o4, %o1, %o4	! 21
+	mulscc	%o4, %o1, %o4	! 22
+	mulscc	%o4, %o1, %o4	! 23
+	mulscc	%o4, %o1, %o4	! 24
+	mulscc	%o4, %o1, %o4	! 25
+	mulscc	%o4, %o1, %o4	! 26
+	mulscc	%o4, %o1, %o4	! 27
+	mulscc	%o4, %o1, %o4	! 28
+	mulscc	%o4, %o1, %o4	! 29
+	mulscc	%o4, %o1, %o4	! 30
+	mulscc	%o4, %o1, %o4	! 31
+	mulscc	%o4, %o1, %o4	! 32
+	mulscc	%o4, %g0, %o4	! final shift
+
+
+	/*
+	 * Normally, with the shift-and-add approach, if both numbers are
+	 * positive you get the correct result.  With 32-bit two's-complement
+	 * numbers, -x is represented as
+	 *
+	 *		  x		    32
+	 *	( 2  -  ------ ) mod 2  *  2
+	 *		   32
+	 *		  2
+	 *
+	 * (the `mod 2' subtracts 1 from 1.bbbb).  To avoid lots of 2^32s,
+	 * we can treat this as if the radix point were just to the left
+	 * of the sign bit (multiply by 2^32), and get
+	 *
+	 *	-x  =  (2 - x) mod 2
+	 *
+	 * Then, ignoring the `mod 2's for convenience:
+	 *
+	 *   x *  y	= xy
+	 *  -x *  y	= 2y - xy
+	 *   x * -y	= 2x - xy
+	 *  -x * -y	= 4 - 2x - 2y + xy
+	 *
+	 * For signed multiplies, we subtract (x << 32) from the partial
+	 * product to fix this problem for negative multipliers (see mul.s).
+	 * Because of the way the shift into the partial product is calculated
+	 * (N xor V), this term is automatically removed for the multiplicand,
+	 * so we don't have to adjust.
+	 *
+	 * But for unsigned multiplies, the high order bit wasn't a sign bit,
+	 * and the correction is wrong.  So for unsigned multiplies where the
+	 * high order bit is one, we end up with xy - (y << 32).  To fix it
+	 * we add y << 32.
+	 */
+#if 0
+	tst	%o1
+	bl,a	1f		! if %o1 < 0 (high order bit = 1),
+	add	%o4, %o0, %o4	! %o4 += %o0 (add y to upper half)
+1:	rd	%y, %o0		! get lower half of product
+	retl
+	addcc	%o4, %g0, %o1	! put upper half in place and set Z for %o1==0
+#else
+	/* Faster code from tege@sics.se.  */
+	sra	%o1, 31, %o2	! make mask from sign bit
+	and	%o0, %o2, %o2	! %o2 = 0 or %o0, depending on sign of %o1
+	rd	%y, %o0		! get lower half of product
+	retl
+	addcc	%o4, %o2, %o1	! add compensation and put upper half in place
+#endif
+
+Lmul_shortway:
+	/*
+	 * Short multiply.  12 steps, followed by a final shift step.
+	 * The resulting bits are off by 12 and (32-12) = 20 bit positions,
+	 * but there is no problem with %o0 being negative (unlike above),
+	 * and overflow is impossible (the answer is at most 24 bits long).
+	 */
+	mulscc	%o4, %o1, %o4	! 1
+	mulscc	%o4, %o1, %o4	! 2
+	mulscc	%o4, %o1, %o4	! 3
+	mulscc	%o4, %o1, %o4	! 4
+	mulscc	%o4, %o1, %o4	! 5
+	mulscc	%o4, %o1, %o4	! 6
+	mulscc	%o4, %o1, %o4	! 7
+	mulscc	%o4, %o1, %o4	! 8
+	mulscc	%o4, %o1, %o4	! 9
+	mulscc	%o4, %o1, %o4	! 10
+	mulscc	%o4, %o1, %o4	! 11
+	mulscc	%o4, %o1, %o4	! 12
+	mulscc	%o4, %g0, %o4	! final shift
+
+	/*
+	 * %o4 has 20 of the bits that should be in the result; %y has
+	 * the bottom 12 (as %y's top 12).  That is:
+	 *
+	 *	  %o4		    %y
+	 * +----------------+----------------+
+	 * | -12- |   -20-  | -12- |   -20-  |
+	 * +------(---------+------)---------+
+	 *	   -----result-----
+	 *
+	 * The 12 bits of %o4 left of the `result' area are all zero;
+	 * in fact, all top 20 bits of %o4 are zero.
+	 */
+
+	rd	%y, %o5
+	sll	%o4, 12, %o0	! shift middle bits left 12
+	srl	%o5, 20, %o5	! shift low bits right 20
+	or	%o5, %o0, %o0
+	retl
+	addcc	%g0, %g0, %o1	! %o1 = zero, and set Z